A Mixture Record Linkage Approach for US Patent Inventor Disambiguation

摘要

Inventor name disambiguation is a task that distinguishes each unique inventor from all other inventor records in patent database. This task is essential for processing person name queries in order to get information related to certain inventor. We proposed a mixture approach that applies to the combination of supervised learning, stochastic record linkage and ruled-based method to determine whether each pair of inventor records are from same inventor or not. Our algorithm tested on the USPTO patent database disam-biguated 12 million inventor records in 7 h. Evaluation is on labeled dataset from USPTO PatentsView inventor name disambiguation competition and showed our approach have an excellent output.